地震数据具有高维特性,而现有的地震数据去噪方法难以处理高维空间的非线性模式数据,如地震剖面的弯曲同相轴。为此,提出利用局部线性嵌入(LLE)和主成分分析(PCA)方法对含有非线性模式的地震数据进行去噪处理。首先,利用LLE重构方式对地震图像采样点用其近邻进行重建;然后,利用PCA分解对LLE重构后的地震图像进行有效信号和噪声分离,去除不相关的噪声。最后,在正演模型和真实地震资料上的实验结果表明,提出的方法有效地消除了随机噪声。 However, the existing seismic data denoising method is difficult to deal with the non-linear model data of high-dimensional space, such as the bending phase of the seismic profile. Therefore, it is proposed to use the local linear embedding (LLE) and principal component analysis (PCA) to denoise the seismic data with nonlinear model. Firstly, the LLE reconstruction method is used to reconstruct the seismic image samples with its neighbors. Then, PCA decomposition is used to separate the effective signal and noise of the LLE reconstructed seismic image to remove the irrelevant noise. Finally, the experimental results on forward modeling and real seismic data show that the proposed method effectively eliminates random noise.